Microelectromechanical structures defined on silicon on insulator wafers are described. More particularly, micromachined structures monolithically integrated with CMOS circuitry are described.
Inherent thin film properties of materials limit many surface micromachining processes. For example, variability of materials properties in polysilicon thin films (such as, for example, Young""s modulus and Poisson""s ratio, residual stress, and stress gradients) can prohibit manufacture of desired microstructures. This is particularly apparent in microoptical components such as mirrors, lenses, and diffraction gratings, which must be very flat for high-optical performance, and normally have to be made in the single crystal silicon layer. Since conventional surface micromachining requires that all components be made in polysilicon layers, optical performance can be limited.
The leading commercial microelectromechanical (MEMS) processing technologies are 1) bulk micromachining of single crystal silicon, and 2) surface micromachining of polycrystalline silicon. Each of these processing technologies has associated benefits and barriers. Bulk micromachining of single crystal silicon, an excellent material with well controlled electrical and mechanical properties in its pure state, has historically utilized wet anisotropic wet etching to form mechanical elements. In this process, the etch rate is dependent on the crystallographic planes that are exposed to the etch solution, so that mechanical elements are formed that are aligned to the rate limiting crystallographic planes. For silicon these planes are the (1,1,1) crystal planes. The alignment of mechanical features to the crystallographic planes leads to limitations in the geometries that can be generated using this technique. Typical geometries include v-groove trenches and inverted pyramidal structures in (1,0,0) oriented silicon wafers, where the trenches and inverted pyramids are bound by (1,1,1) crystallographic planes. Geometries that include convex corners are not allowed unless additional measures are taken to protect etching of the crystal planes that make up the corners. The etch rate also varies with dopant concentration, so that the etch rate can be modified by the incorporation of dopant atoms, which substitute for silicon atoms in the crystal lattice. A boron dopant concentration on the order of 5xc3x971019/cm3 is sufficient to completely stop etching, so that mechanical elements bounded by other crystal planes can be generated by using dopant xe2x80x9cetch stopxe2x80x9d techniques. However dopant concentrations of this magnitude are sufficient to modify the desirable electrical and mechanical properties of the pure single crystal silicon material, leading to device design and manufacturability constraints. Recent advances in Deep Reactive Ion Etching (DRIE) (J. K. Bhardwaj and H. Ashraf, xe2x80x9cAdvanced silicon etching using high density plasmasxe2x80x9d, Micromachining and Microfabrication Process Technology, Oct. 23-24, 1995, Austin, Tex., SPIE Proceedings Vol. 2639, pg. 224) which utilizes sidewall passivation and ion beam directionality to achieve etch anisotropy, has relaxed the in-plane geometric design constraints, but still require etch stop techniques to control the depth of the etch into the wafer, and additional processing steps are required to undercut a structure to release it from the substrate.
In contrast, surface micromachining of polycrystalline silicon utilizes chemical vapor deposition (CVD) and reactive ion etching (RIE) patterning techniques to form mechanical elements from stacked layers of thin films (R. T. Howe, xe2x80x9cSurface micromachining for microsensors and microactuatorsxe2x80x9d, J. Vac. Sci. Technol. B6, (1988) 1809). Typically CVD polysilicon is used to form the mechanical elements, CVD nitride is used to form electrical insulators, and CVD oxide is used as a sacrificial layer. Removal of the oxide by wet or dry etching releases the polysilicon thin film structures. The advantage of the surface micromachining process is the ability to make complex structures in the direction normal to the wafer surface by stacking releasable polysilicon layers (K. S. J. Pister, M. W. Judy, S. R. Burgett, and R. S. Fearing, xe2x80x9cMicrofabricated hingesxe2x80x9d, Sensors and Actuators A33, (1992) 249 and L. Y. Lin, S. S. Lee, K. S. J. Pister, and M. C. Wu, xe2x80x9cMicromachined three-dimensional micro-optics for free-space optical systemxe2x80x9d, IEEE Photon. Technol. Lett. 6, (1994) 1445) and complete geometric design freedom in the plane of the wafer since the device layers are patterned using isotropic RIE etching techniques. An additional advantage of surface micromachining is that it utilizes thin film materials such as polysilicon, oxide, nitride, and aluminum, that are commonly used in microelectronic device fabrication, albeit with different materials properties that are optimized for mechanical rather than electrical performance. This commonality in materials allows for increased integration of microelectronic and micromechanical components into the same fabrication process, as demonstrated in Analog Devices"" integrated accelerometer, and in SSI Technologies"" integrated pressure sensor.
While surface micromachining relaxes many of the limitations inherent in bulk micromachining of single crystal silicon, it nonetheless has its own limitations in thin film properties. The maximum film thickness that can be deposited from CVD techniques are limited to several microns, so that thicker structures must be built up from sequential depositions. Thicker device layers are required for dynamic optical elements where dynamic deformations can impact optical performance, and for optical elements which require additional thin film coatings that can cause stress induced curvature. The thin film mechanical properties, such as Young""s modulus and Poisson""s ratio, are dependent on the processing parameters and the thermal history of the fabrication process, and can typically vary by as much as 10% from run to run. This is an important limitation for robust manufacturability where these thin film mechanical properties can be a critical parameter for device performance. An additional limitation of conventional surface micromachining is that holes through the mechanical elements must be included in the design to allow the etchants used to release the mechanical elements to reach the sacrificial layers. While this is not an important limitation for optical elements such as Fresnel lenses and diffraction grating that include holes in their design, it is an important limitation for optical elements such as mirrors where holes are a detriment to optical performance. Flatness and reflectivity are also important optical design criteria that can be impacted by conventional surface micromachining processes. Thin film stresses and stress gradients, typical of polysilicon thin films, can lead to warping of optical surfaces. In addition the surface of as-deposited polysilicon thin films is not polished, and thus requires post-processing Chemical Mechanical Polishing (CMP) techniques to obtain an optical quality surface finish.
Since the fabrication technology utilized to produce microoptoelectromechanical (MOEMS) components can lead to manufacturing barriers in the thin film properties associated with the process, the present invention includes an enabling fabrication process for microoptoelectromechanical systems that overcomes the barriers in the optomechanical properties of thin film structures. The key innovation to overcoming these thin film properties is to utilize silicon on insulator (SOI) wafers as the starting substrate in a surface micromachining process (see FIG. 1). SOI is a generic term that refers to a structure in which a silicon layer is supported by a dielectric material. In this embodiment, a silicon device layer, bonded to a conventional silicon handle wafer, has a SiO2 thin-film layer at the interface. This allows critical optical and electronic components to be fabricated in a single crystal silicon device layer, which can be released from the handle wafer by etching the oxide at the interface between the device layer and the substrate. The oxide layer at the interface can also be utilized as a backside etch stop layer for releasing optical components, such as a mirror, that cannot include etch holes. The device layer has a user specified thickness that is appropriate for the given application, and has excellent and reproducible electrical and thin film properties. Both the back and front side of the device layer would be polished, and thus optical elements fabricated in this layer do not require additional post-processing CMP techniques to obtain an optical quality surface finish. Since the device layer is single crystal silicon, it has no intrinsic stress or stress gradients in the absence of thin film coatings. Since it can be made thicker than conventional CVD deposited thin films, optical components fabricated in this layer have minimal distortions after thin film depositions such as aluminum to increase surface reflectivity, or dielectric thin films to decrease surface reflectivity. The additional thickness is also important to minimize distortions for dynamically actuated optical elements.
Additional functions, objects, advantages, and features of the present invention will become apparent from consideration of the following description and drawings of preferred embodiments.